web-dev-qa-db-fra.com

Tensorflow: manque de mémoire pour essayer d'allouer 3,90 Go. L'appelant indique qu'il ne s'agit pas d'un échec

Il y a une question que je ne comprends pas. 

Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.90GiB. 
The caller indicates that this is not a failure, 
but may mean that there could be performance gains if more memory is available.

Quelle est la phrase signifie?

J'ai lu le code source. Mais je ne peux pas comprendre à cause de ma faible capacité.
La taille de la mémoire du GPU est de 6 Go, le résultat de l'utilisation de la mémoire par l'analyse tfprof est d'environ 14 Go . Cela dépasse la taille de la mémoire du GPU . mémoire de la CPU ou utiliser le bon algorithme sur l'utilisation de la mémoire du GPU? 

The version of tensorflow that I use is 1.2.

L'infomition de GPU en tant que flux:

  • nom: GeForce GTX TITAN Z
  • majeur: 3 mineurs: 5 memoryClockRate (GHz) 0.8755
  • Mémoire totale: 5,94 Go
  • Mémoire disponible: 5.87GiB

Mon code:

#!/usr/bin/python3.4

import tensorflow as tf
import tensorflow.examples.tutorials.mnist.input_data as input_data
import os
os.environ["CUDA_VISIBLE_DEVICES"] = '0' 


mnist = input_data.read_data_sets("MNIST_data", one_hot=True)
sess = tf.InteractiveSession(config=tf.ConfigProto(log_device_placement=True))
#sess = tf.InteractiveSession()
def weight_variable(shape):
    init = tf.truncated_normal(shape, stddev=0.1)
    return tf.Variable(init)

def bias_variable(shape):
    init = tf.constant(0.1, shape=shape)
    return tf.Variable(init)

def conv2d(x, W):
    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')

def max_pool_2x2(x):
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')

x = tf.placeholder(tf.float32, [None, 784])
y_ = tf.placeholder(tf.float32, [None, 10])
x_image = tf.reshape(x, [-1, 28, 28, 1])


W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)


W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)


W_f1 = weight_variable([7*7*64, 1024])
b_f1 = bias_variable([1024])
h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_f1) + b_f1)


keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

W_f2 = weight_variable([1024, 10])
b_f2 = bias_variable([10])
y_conv = tf.nn.softmax(tf.matmul(h_fc1_drop, W_f2) + b_f2)

cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_conv), reduction_indices=[1]))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))


test_images = tf.placeholder(tf.float32, [None, 784])
test_labels = tf.placeholder(tf.float32, [None, 10])


tf.global_variables_initializer().run()

run_metadata = tf.RunMetadata()


for i in range(100):
    batch = mnist.train.next_batch(10000)
    if (i%10 == 0):  
        train_accurancy = accuracy.eval(feed_dict={x: batch[0], y_: batch[1], keep_prob : 1.0})
        print("step %d, traning accurancy %g" % (i, train_accurancy))
    sess.run(train_step, feed_dict={x: batch[0], y_: batch[1], keep_prob : 0.5}, options=tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE), run_metadata=run_metadata)

tf.contrib.tfprof.model_analyzer.print_model_analysis(
    tf.get_default_graph(),
    run_meta=run_metadata,
    tfprof_options=tf.contrib.tfprof.model_analyzer.PRINT_ALL_TIMING_MEMORY)

test_images = mnist.test.images[0:300, :]
test_labels = mnist.test.labels[0:300, :]
print("test accuracy %g" % accuracy.eval({x: test_images, y_: test_labels, keep_prob: 1.0}))

L'avertissement:

2017-08-10 21:37:44.589635: W tensorflow/core/common_runtime/bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.90GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.
2017-08-10 21:37:46.208897: W tensorflow/core/common_runtime/bfc_allocator.cc:217] Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.61GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory is available.

Le résultat de tfprof:

==================Model Analysis Report======================
_TFProfRoot (0B/14854.97MB, 0us/7.00ms)
tensorflowout-of-memory
5
qihao

vous utilisez le GPU, et votre taille de lot est de 1000. Cela fait beaucoup pour 10 classes! faire une taille de lot plus petite comme 10to20 et augmenter la plage à 10e4 ou même 10e3. Ce problème est bien connu . Si vous voulez vraiment utiliser 10000 comme taille de lot, dites à tensorflow d’utiliser le CPU avec: 

tf.device('/cpu:0')
1
Engine